Active Learning with Body Mass Index Heuristic

This workflow uses a simple example to demonstrate one possible structure for an active learning application and compares the effectiveness of the active learning strategy vs a random labeling approach.

The example model is trained to predict whether a subject on the weight-height plane is underweight or overweight. The heuristic used to provide labels is called body mass index (BMI).

In a real application of the active learning loop, replace the Rule Engine nodes with your method of labeling. For example the Label View node for easy labeling in the KNIME WebPortal.

External resources

• Wikipedia - Body Mass Index
• Burr Settles, Active Learning Literature Survey, 2010 - Chapter 3.1 Uncertainty Sampling
• Active learning for object classification - Nicolas Cebron et al - Data Min Knowl Disc (2009)